def rcnn_resize(im, target_size, max_size, stride=0):
"""
only resize input image to target size and return scale
:param im: BGR image input by opencv
:param target_size: one dimensional size (the short side)
:param max_size: one dimensional max size (the long side)
:param stride: if given, pad the image to designated stride
:return:
"""
im_shape = im.shape
im_size_min = np.min(im_shape[0:2])
im_size_max = np.max(im_shape[0:2])
im_scale = float(target_size) / float(im_size_min)
# prevent bigger axis from being more than max_size:
if np.round(im_scale * im_size_max) > max_size:
im_scale = float(max_size) / float(im_size_max)
im = image.resize(im,
int(im.shape[1] * float(im_scale)),
int(im.shape[0] * float(im_scale)),
interp=1)
if stride == 0:
return im, im_scale
else:
# pad to product of stride
im_height = int(np.ceil(im.shape[0] / float(stride)) * stride)
im_width = int(np.ceil(im.shape[1] / float(stride)) * stride)
im_channel = im.shape[2]
padded_im = np.zeros((im_height, im_width, im_channel))
padded_im[:im.shape[0], :im.shape[1], :] = im
return padded_im, im_scale
def _preprocess(self, data):
input_shape = self.signature['inputs'][0]['data_shape']
height, width = input_shape[2:]
img_arr = image.read(data[0])
img_arr = image.resize(img_arr, width, height)
img_arr = image.color_normalize(img_arr, nd.array([127.5]),
nd.array([127.5]))
img_arr = image.transform_shape(img_arr)
return [img_arr]
def _preprocess(self, data):
img_list = []
for idx, img in enumerate(data):
input_shape = self.signature['inputs'][idx]['data_shape']
# We are assuming input shape is NCHW
[h, w] = input_shape[2:]
img_arr = image.read(img)
img_arr = image.resize(img_arr, w, h)
img_arr = image.transform_shape(img_arr)
img_list.append(img_arr)
return img_list
def _preprocess(self, data):
img_list = []
for idx, img in enumerate(data):
input_shape = self.signature['inputs'][idx][
'data_shape'] # Input shape is NCHW
[h, w] = input_shape[2:]
img_arr = image.read(
img, 0) #Set flag to 0 for reading grayscale images
img_arr = image.resize(img_arr, w, h)
img_arr = image.transform_shape(img_arr)
img_list.append(img_arr)
return img_list
def _preprocess(self, data):
img_list = []
for idx, img in enumerate(data):
input_shape = self.signature['inputs'][idx]['data_shape']
# We are assuming input shape is NCHW
[h, w] = input_shape[2:]
if input_shape[1] == 1:
img_arr = image.read(img, 0)
else:
img_arr = image.read(img)
img_arr = image.resize(img_arr, w, h)
rgb_mean = mx.nd.array([123.68, 116.779, 103.939])
rgb_std = mx.nd.array([58.395, 57.12, 57.375])
img_arr = img_arr.astype('float32')
img_arr = mx.image.color_normalize(img_arr,
mean=rgb_mean,
std=rgb_std)
img_arr = mx.nd.reshape(img_arr, (w, h, input_shape[1]))
img_arr = image.transform_shape(img_arr)
img_list.append(img_arr)
return img_list
def test_resize(self):
input1 = mx.nd.random.uniform(0, 255, shape=(245, 156, 3))
output1 = image.resize(input1, 128, 256)
assert output1.shape == (256, 128, 3), "Resize method failed. Got %s shape." % (str(output1.shape))